Research On The Applications Of Wavelet Transform To Image Edge Extraction And Image Coding

Wavelet transform is a new analysis theory developed in last two decades. The applications of wavelet analysis to image processing is the topic of this thesis. The main contributions in this paper can be summarized as follows:?The choice of the threshold in Mallat抯 image multiscale edges extraction based on wavelet transfonnation is improved. A new way of choosing the threshold by using rectangle self-adapting method is presented.?A family of biorthonormal wavelets suitable for multiresolution edge extraction is constructed. Experiment results are given to show its correctness and effectiveness.?An improved adaptive zerotree based embedded image coding of wavelet coefficients is proposed.With our 揷ommon K-L transformation?method, more zerotrees will be produced so that the PSNR can be improved by 0.1 ?.6dB at the same bit rate as the original method. Furthermore, the code stream produced is also embedded and able to support multirate image decoding as the original method.?We justify the applicability of antisymmetry biorthonormal wavelets (ASBW) to image coding by a number of experiments, and found that the performance of some selected ASBWs is at least the same as the famous 9₇ wavelet, which is a symmetry biorthonormal wavelet. However ASBW has a potential advantage, namely extracting image edges from the coefficients decomposed by ASBW, it may give a way to feature extraction from image compression data domain without completely image reconstruction.